In this paper, we introduce a domain-specific tool designed to streamline the development of deep learning models for industrial applications, particularly those facing the challenges posed by imbalanced datasets. The advent of pervasive computing has allowed the creation of new smart services based on machine learning, yet the complexity of effectively harnessing data from IoT devices remains a challenge. The data characteristics - ranging from noise and scarcity to a lack of labels and imbalance - present significant obstacles for their utilization. A focal point of this research is the issue of class imbalance, which significantly impedes the achievement of accurate predictive models, especially in critical applications like anomaly detection. Our paper presents a solution through the development of a specialized tool that integrates state-of-the-art techniques and metrics specifically designed to deal with class imbalance. This tool also adapts to various constraints set forth by domain experts, thereby facilitating the creation of optimized models suited to specific industrial needs. To demonstrate its practicality and effectiveness, we apply our tool in a case study focusing on the detection of motor malfunctions in the context of Industry 4.0, in collaboration with Schneider Electric.

The use of supervised learning for Human Activity Recognition (HAR) on mobile devices leads to strong classification performances. Such an approach, however, requires large amounts of labeled data, both for the initial training of the models and for their customization on specific clients (whose data often differ greatly from the training data). This is actually impractical to obtain due to the costs, intrusiveness, and time-consuming nature of data annotation. Moreover, even with the help of a significant amount of labeled data, model deployment on heterogeneous clients faces difficulties in generalizing well on unseen data. Other domains, like Computer Vision or Natural Language Processing, have proposed the notion of pre-trained models, leveraging large corpora, to reduce the need for annotated data and better manage heterogeneity. This promising approach has not been implemented in the HAR domain so far because of the lack of public datasets of sufficient size. In this paper, we propose a novel strategy to combine publicly available datasets with the goal of learning a generalized HAR model that can be fine-tuned using a limited amount of labeled data on an unseen target domain. Our experimental evaluation, which includes experimenting with different state-of-the-art neural network architectures, shows that combining public datasets can significantly reduce the number of labeled samples required to achieve satisfactory performance on an unseen target domain.

Federated Learning has been introduced as a new machine learning paradigm enhancing the use of local devices. At a server level, FL regularly aggregates models learned locally on distributed clients to obtain a more general model. Current solutions rely on the availability of large amounts of stored data at the client side in order to fine-tune the models sent by the server. Such setting is not realistic in mobile pervasive computing where data storage must be kept low and data characteristic can change dramatically. To account for this variability, a solution is to use the data regularly collected by the client to progressively adapt the received model. But such naive approach exposes clients to the well-known problem of catastrophic forgetting. To address this problem, we have defined a Federated Continual Learning approach which is mainly based on distillation. Our approach allows a better use of resources, eliminating the need to retrain from scratch at the arrival of new data and reducing memory usage by limiting the amount of data to be stored. This proposal has been evaluated in the Human Activity Recognition (HAR) domain and has shown to effectively reduce the catastrophic forgetting effect.

Pervasive computing promotes the installation of connected devices in our living spaces in order to provide services. Two major developments have gained significant momentum recently: an advanced use of edge resources and the integration of machine learning techniques for engineering applications. This evolution raises major challenges, in particular related to the appropriate distribution of computing elements along an edge-to-cloud continuum. About this, Federated Learning has been recently proposed for distributed model training in the edge. The principle of this approach is to aggregate models learned on distributed clients in order to obtain a new, more general model. The resulting model is then redistributed to clients for further training. To date, the most popular federated learning algorithm uses coordinate-wise averaging of the model parameters for aggregation. However, it has been shown that this method is not adapted in heterogeneous environments where data is not identically and independently distributed (non-iid). This corresponds directly to some pervasive computing scenarios where heterogeneity of devices and users challenges machine learning with the double objective of generalization and personalization. In this paper, we propose a novel aggregation algorithm, termed FedDist, which is able to modify its model architecture (here, deep neural network) by identifying dissimilarities between specific neurons amongst the clients. This permits to account for clients' specificity without impairing generalization. Furthermore, we define a complete method to evaluate federated learning in a realistic way taking generalization and personalization into account. Using this method, FedDist is extensively tested and compared with three state-of-the-art federated learning algorithms on the pervasive domain of Human Activity Recognition with smartphones.

Pervasive computing allows users to benefit from information and services provided by advanced applications using smart and interconnected devices. Given the inherent complexity of this software environment, platforms become a typical means of easing the development and execution of pervasive applications. These platforms are usually in charge of managing the application context, ensuring that data are reliable and available to applications. Most platforms provide this feature, making real-time information available, even when applications do not need such information in real-time. This occurs, for example, when sensors continuously provide measurements to in-house temperature controllers and these measurements have limited time variance. Therefore, controllers can be updated only occasionally, reducing application overload without compromising effectiveness. To address this issue, in this paper, we present a pervasive platform, iCasa, extended with an autonomic manager of service-oriented context module. Our solution consists of tracking application needs and context information to achieve the best caching configuration, regarding a multi-application pervasive environment. This caching configuration can decrease the power consumption of devices due to the reduced unnecessary network communications between platform and devices.

Pervasive computing vision considers the growing concentration of small computers merged into everyday objects to enhance environment with added-value services. In this vision computing systems are numerous, heterogeneous, and dynamic, yet seamlessly integrated. The complexity of pervasive environments can lead to conflicts between systems, but it can also be a strength if systems manage to collaborate. They need to understand and reason about their objectives, capabilities and environment. This paper investigates the definition and the relevancy of self-awareness for leveraging collaboration in pervasive environment, with a model-driven approach.

Healthcare costs are booming as population ages and need healthcare. Pervasive applications are expected to take healthcare into people's homes and thus cut health-related costs. However, the lack of appropriate middleware to collect and integrate pervasive data is a brake to their development. In this paper, we present the integration of iCasa and Cilia, two service-oriented frameworks, and show it can be applied to medical applications. These frameworks facilitate the work of developers and exhibit autonomic features easing administration.

The purpose of pervasive applications is to be invisible and to require a minimal amount of user attention. But invisibility raises lots of hard challenges, blocking the long-awaited pervasive era to emerge. The device heterogeneity and volatility, security enforcement and context-awareness are just some of the issues a developer must tackle inside the application code. To industrialize such development, we propose to use an application server dedicated to the pervasive environment. Thanks to this server, developers focus on the application, delegating to the server most of the pervasive constraint management. However, providing such a platform is a very difficult task. The pervasive environment constraints must be handled by this server in a consistent way, while being flexible enough to support the different applications. This paper describes the requirements of the pervasive applications and how our platform, named iCASA, deals with them.

In this paper, we present a distributed and autonomic mediation framework named Cilia, dedicated to complex physical environments like manufacturing plants, electrical infrastructures or smart buildings. We detail a motivating example in these fields and we explain how autonomic features are absolutely necessary in domains characterized by important environmental dynamism and by evolving needs of decision-makers. We also present in detail the main principles of Cilia, including its autonomic abilities. We believe that these autonomic behaviors are necessary to deal with Big Data in industrial contexts. Cilia has been validated in collaborative projects with Orange and Schneider Electric.

Autonomic computing seeks to render computing systems as self-managed. In other words, its objective is to enable computer systems to manage themselves so as to minimise the need for human input. Software architectures can be used in order to express constraints to be maintained all along the execution of a system and, conversely, to present the state of the running system. In this paper, we present an approach where design and runtime architectures are used to manage service-oriented systems. We show how concepts of design time and runtime can be linked and exploited by an autonomic manager or by a human administrator. This approach is validated on a real use case belonging to the pervasive domain.

In the last years, we developed web applications requiring a large amount of JavaScript code. These web applications present adaptation requirements. In addition to platform-centric adaptation, applications have to dynamically react to external events like connectivity disruptions. Building such applications is complex and we faced sharp maintainability challenges. This paper presents h-ubu, a service-oriented component framework for JavaScript allowing building adaptive applications. h-ubu is used in industrial web applications and mobile applications. h-ubu is available in open source, as part of the OW2 Nanoko project.

In the context of pervasive environments, multimodal interaction plays a pivotal role because multimodality provides flexibility and naturalness for interaction. The challenge of multimodal interfaces in pervasive environments is then to build reliable and autonomic processing systems able to analyze and understand multiple interaction modalities and reconfigure itself in real-time. Addressing this issue, we have developed an autonomic framework called DynaMo (Dynamic multiMOdality) for the development and runtime management of multimodal interaction in pervasive environments. DynaMo is composed by a specification language dedicated to the multimodality domain and a runtime machine that instantiates these specifications. In this paper, we present the overall architecture of our solution DynaMo that is based on partial interaction models, and how these models are completed at runtime to build multimodal interfaces adapted to the local execution environment.

In this paper, we describe the work that has been realized in order to add autonomic features to Cilia, an open source mediation framework developed jointly by the LIG/Adele team at Grenoble University and Orange Labs. Cilia has been designed for data and application mediation and is used in several industrial use cases. This paper, in particular, develops the notion of state variables and action variables adapted from control theory. It shows how they can be used to follow the state of integration chains and to bring changes at runtime.

Heterogeneity and dynamicity of pervasive environments require the construction of flexible multimodal interfaces at run time. In this paper, we present how we use an autonomic approach to build and maintain adaptable input multimodal interfaces in smart building environments. We have developed an autonomic solution relying on partial interaction models specified by interaction designers and developers. The role of the autonomic manager is to build complete interaction techniques based on runtime conditions and in conformity with the predicted models. The sole purpose here is to combine and complete partial models in order to obtain an appropriate multimodal interface. We illustrate our autonomic solution by considering a running example based on an existing application and several input devices.

This article introduces the design of the AutoHome service-oriented framework to simplify the development and runtime adaptive support of autonomic pervasive applications. To this end, we describe our novel open infrastructure for building and executing home applications. This includes the amalgamation of the two computing areas of autonomics and service orientation, to produce a component-based platform providing facilities including monitoring, touchpoints, and other common autonomic services. This infrastructure uniquely blends the advantages of distributed autonomic control with global conflict management in a management hierarchy. We discuss this platform in terms of pervasive home systems and show how one would develop such a system for two examples of automated home applications: intruder detection and medical support, respectively. Both applications were built within our framework and evaluated showing that the use of the framework introduces minimal overheads but provides many benefits. We then conclude by highlighting the contributions of AutoHome and a discussion about the lessons learned, limitations, and future research directions.

The recent evolution of mobile smart devices and their convergence with pervasive computing and software as a service raises new challenges. Developers of applications targeted to these environments have to face at least three major challenges: dynamicity, heterogeneity, and distribution. In this paper, we propose a service oriented component framework which addresses these challenges by automatically reifying available services in a distributed pervasive environment. Therefore, the only thing developers have to focus on is writing the business code. We have implemented and validated our framework by using several real applications developed within collaborative projects (including ITEA ANSO project).

Since recently, it is accepted that mediation can take many forms depending on the application domains: for example, mediation can be seen as a means to deal with interoperability between legacy applications in large Information Systems in a variety of environments including Cloud or to synchronize, integrate and route data from sensor networks in pervasive applications. In this paper we present a lightweight, modular, dynamic and distributed framework called Cilia dedicated to mediation. Cilia aims at bringing an homogeneous solution to deal with mediation both in big and little systems and to ensure the seamless integration of the two systems. In this paper we demonstrate the benefits of the Cilia approach through a case study from an RFID project.

Software Product Lines (SPLs) engineering aims to develop similar software systems by sharing a common set of software assets. Service-Oriented Computing (SOC) uses services as fundamental elements for developing applications. SOC provide dynamic capabilities that are needed in many product-lines. SPLs provide useful mechanisms for modeling dynamic applications implemented through service compositions. The main contribution of this paper is to propose a three-phase approach that consolidates the two paradigms for developing dynamic service compositions. The proposed approach is supported by an integrated development environment including three tools. The first tool allows specifying product line architecture integrating variability model. The second tool allows defining service compositions via the assembly of service descriptions. Such service compositions are designed following the product line architecture. Executing applications made of service instances can be autonomously built by a runtime infrastructure. The paper exemplifies the use of the presented tools in the home hospitalization domain.

In this paper, we propose an approach based on Formal Concept Analysis in order to organize the services registry at runtime and to allow the ”best” service selection among heterogeneous and secured services according to a set of specifications. This solution has been validated in the European SODA project on pervasive applications.

Service composition is complex. It has to reach a set of pre-defined non-functional qualities, like security for instance, which requires the production of complicated code.This code, often distributed between client and server sides, is highly error-prone and difficult to maintain. In this paper, we present a generative environment for the orchestration of abstract services and the separate specification of non-functional properties. This environment has been built within the European SODA project and validated on several industrial use cases. In this paper, we focus on an alarm management scenario with stringent security requirements.

Computer aided software engineering tools represent one the main successes of software engineering in the past decades. They however need to be improved along several dimensions in order to face new challenges due to ever more complex applications, more heterogeneous technologies and more stakeholders involved. In this paper, we present an approach based on the concept of domain. We define a domain as an area in which a number of stakeholders is repeatedly performing similar activities. In a project, an arbitrary number of domains can be identified, being business, technical, or related to life cycle activities. In our metamodel-based approach, any domain can be easily modelled and the corresponding computer aided domain specific engineering environment (CADSE) can be generated. Using CADSE composition, complete and wide scope engineering environments can be built as a composition of an arbitrary number of domains. The paper presents the approach, the technology and draws a few lessons of the first years of use in a number of real projects.

The emergence of smart devices, especially in the home context, will change the way we interact with our environment. However, implementing and maintaining home-control applications is still far from easy. This paper discusses the difficulties encountered during the development of a home gateway. It then introduces a framework that sets the basics of a home application server and allows an easy development of home applications. An implementation is subsequently presented to validate the approach and its benefits.

The home environment becomes ready to host distributed devices dynamically adapting to service availability and reacting to user location and user activity. Sensors, high definition rendering systems, home gateways, wired and wireless controllable equipments are now available. Many protocols enable connectivity and interaction between devices. However, challenges remain: protocol heterogeneity, interface fragmentation and device composition static aspect make self-organization and dynamic reconfiguration hardly achievable. This paper describes attractive scenarios at home which lead to the definition of the pervasive service composition requirements. A software architecture facing the mentioned challenges is proposed over OSGi. It first enables developers to implement distributed plug-n-play applications like a local one. It also delivers a service-oriented middleware allowing spontaneous distributed service composition to occur at runtime.

A broad range of industries must be able to seamlessly integrate control applications using sensors distributed in the real world and IT applications supporting various business activities. Seamless integration requires building distributed systems in distributed, heterogeneous execution environments. In order to meet these requirements, innovative service- oriented architectures have been recently adopted by major manufacturers. This service-oriented architecture raises numerous issues, including deployment of service-oriented applications. In this paper, we propose an adapted set of activities expressing the deployment process, a service application description language (via a meta-model) which is technologies free and a deployment manager taking the applications models as input and putting in place the different activities.

Abstract — The convergence of smart field devices and business services stands to profoundly change the way we interact with our environment. However, implementing and maintaining home-control applications is still far from easy. This paper discusses how service-oriented concepts can facilitate the development of home-control applications. Moreover we present an open architecture to implement these applications. A prototype of service platform for home-control is presented to illustrate the approach and its benefits. Service oriented computing, home control applications, service platform, OSGi™, iPOJO, application model. I.

The convergence of smart home appliances and business services stands to profoundly change the way we interact with our environment. However, implementing and maintaining home-control applications is still far from easy. This demonstration presents our service-oriented platform to facilitate the development of home-control applications. This prototype is used in the ANSO industrial European project.

A good software architecture facilitates application system development, promotes achievement of functional requirements, and supports system reconfiguration. We present a domain-specific software architecture (DSSA) that we have developed for a large application domain of adaptive intelligent systems (AISs). The DSSA provides: (a) an AIS reference architecture designed to meet the functional requirements shared by applications in this domain, (b) principles for decomposing expertise into highly reusable components, and (c) an application configuration method for selecting relevant components from a library and automatically configuring instances of those components in an instance of the architecture. The AIS reference architecture incorporates features of layered, pipe and filter, and blackboard style architectures. We describe three studies demonstrating the utility of our architecture in the subdomain of mobile office robots and identify software engineering principles embodied in the architecture.>